.png%3Ftable%3Dblock%26id%3D6246082f-4014-4d06-98ab-59e9840b298a%26cache%3Dv2&w=3840&q=75)
We have long thought within a two-dimensional world.
Almost every medium that digital civilization has offered for thought — letters, images, documents, screens — has existed on a flat plane. Space has always existed, but more as an output of thought rather than a language of thought itself. To express something as spatial, you had to build a physical object or master 3D modeling techniques through years of training. The cost was so high that most people simply couldn’t “write” space. Space may have been imagined, but in most cases it was never truly documented.
At this point, AR/VR and the metaverse offered answers far too early. They assumed that once people stepped into virtual space, new kinds of thinking and behavior would naturally follow. But for most, space remained a consumable object, not a medium of thought. It could be viewed, but not written.
The more fundamental reason is that there simply hasn’t been enough material to write with in space. Because 3D modeling and creation are still expensive and difficult, the basic building blocks you can place and combine in space are scarce. Not only finished assets but even reusable templates or modular structures are limited. As a result, space has remained a venue for consuming a small set of content — shown more often, but thought about no more deeply.
Whenever thinking has dramatically evolved, it has been because the cost of externalizing thought dropped sharply. When writing first emerged, humans didn’t suddenly learn new thoughts — they were finally able to accumulate thoughts that were previously possible but unsustainable. 3D sits in the same place. We understand space, but the ability to record it, manipulate it, and share it has long been confined to a small group of experts.
The Arrival of an Ability to “Write” Space with 3D Generative AI
The emergence of 3D generative AI changes this. It is not a technology that simply displays space more realistically. It is a technology that lowers the cost of thinking about space. Spatial relationships — size, distance, movement, arrangement — no longer have to remain only in your head. They can be placed in the external world as easily as speaking words.
The moment space becomes easy to write, people won’t just come up with more ideas — they will start discarding bad ones faster. An idea that sounded plausible in words suddenly feels awkward when placed in space. Thought progress has always begun not by finding the right answer but by quickly rejecting the wrong ones.
This change won’t happen for everyone all at once. Just as not everyone writes even in a literate world, not everyone will create spatial content — but a mindset of explaining and validating through space will first take root in certain domains.
Games: Where the Space You Write Becomes the Result
The first place this change will manifest is games.
In games, the space you write isn’t just a representation — it’s the actual result. What is correct isn’t decided by explanation or agreement — it’s revealed the moment you interact with it.
In real professional environments, many validations happen before play. When designing a massive mechanical structure, artists don’t start by creating a complete form. They define parts at the image level, translate them into 3D, assemble them, and quickly check proportions, structures, and overall silhouettes. If something feels off once assembled, the parts are redesigned and reassembled. These iterative loops happen before play, yet they are clear acts of judgment.
Play is the most intense form of validation. The moment of “actually moving and interacting” transforms assumptions about space from abstract debates into tangible decisions. Thus many ideas are filtered at once — separating what’s worth keeping from what’s not — even before the final form is completed.
Across parts combination, structural experiments, and play — different methods reveal the same pattern. The moment space is placed outside the mind, assumptions become objects of judgment. Writing space isn’t about completing a finished product; it’s about creating a state where you can judge whether your current thinking is right or wrong.
On the Limits of Completion
At the same time, the quality of the generated 3D itself remains another axis of challenge. Current 3D generative AI still shows definite limitations in mesh and texture resolution, topology stability, UV coherence, and reliable PBR material representation
_1.png%3Ftable%3Dblock%26id%3D2d698458-afb5-808d-a163-ceb9c11edeba%26cache%3Dv2&w=3840&q=75)
In many cases, the quality of generated meshes or textures still falls short of what real production pipelines require. UV layouts are often excessively fragmented at the face level, making subsequent editing and reuse difficult, and material representations frequently lack physical plausibility in a real-world context.
In addition, generating consistent rigging and animation across both humanoid and non-humanoid forms remains a challenging problem. Estimating appropriate joint structures, ensuring stable skinning weights, and achieving natural animation all involve requirements that vary significantly depending on an object’s shape and intended use. As a result, developing a truly generalizable solution is far from trivial.
However, the nature of these problems is relatively clear. They are less about demanding a new way of thinking and more about well-defined technical difficulty. AI-driven intelligent retopology, automated UV unwrapping, PBR texture generation, and the generalization of AI-based rigging and animation are already recognized across the industry as shared challenges, and progress in these areas is advancing rapidly. While each stage may still feel incomplete today, the overall direction toward solutions is comparatively well defined.
What matters is that these two issues are not in competition, but exist as parallel axes. The ability to write space expands the dimension of thought, while fidelity and completeness accumulate naturally on top of it. Technology will ultimately be required to advance along both axes simultaneously, and real value in production environments will only emerge when the two move forward together.
The Future of the Ability to Write Space
The ability to write space is not a technology that produces results on your behalf. It is closer to a technology that allows thoughts to be externalized and quickly tested for whether they are right or wrong. Through the repetition of this judgment, implementation and fidelity follow naturally. To say that space becomes a language of thought means that thinking and making are no longer separate processes, but part of a single, continuous flow.
The future of 3D generative AI will unfold along this flow. Not toward automatically producing more things, but toward enabling more ideas to be placed into space — and discarded — with ease. And I believe this transformation cannot reach any other domain without first passing through games, because they demand the most frequent and the most unforgiving judgments before completion.
Games are the first market — and the first sentence.